Impact of a large-scale telemedicine network on emergency visits and hospital admissions during the coronavirus disease 2019 pandemic in Brazil: Data from the UNIMED-BH system

Abstract

Introduction

Triage by on-demand telemedicine is a strategy for healthcare surge control in the COVID-19 pandemic. We aimed to assess the impact of a large-scale COVID-19 telemedicine system on emergency department (ED) visits and all-cause and cardiovascular hospital admissions in Brazil.

Methods

From March 18, 2020–May 18, 2020 we evaluated the database of a cooperative private health insurance, with 1.28 million clients. The COVID-19 telemedicine system consisted of: a) mobile app, which redirects to teleconsultations if indicated; b) telemonitoring system, with regular phone calls to suspected/confirmed COVID-19 cases to monitor progression; c) emergency ambulance system (EAS), with internet phone triage and counselling. ED visits and hospital admissions were recorded, with diagnoses assessed by the Diagnosis Related Groups method. COVID-19 diagnosis and deaths were identified from the patients’ registries, and outcomes assessed until June 1st.

Results

In 60 days, 24,354 patients accessed one of the telemedicine systems. The most frequently utilized was telemonitoring (16,717, 69%), followed by teleconsultation (13,357, 55%) and EAS (687, 3%). The rates of ED and hospital admissions were: telemonitoring 19.7% (3,296) and 4.7% (782); teleconsultation 17.3% (2,313) and 2.4% (318) and EAS: 55.9% (384) and 56.5% (388) patients. At total 4.1% (1,010) had hospital admissions, 36% (363) with respiratory diseases (44 requiring mechanical ventilation) and 4.4% (44) with cardiovascular diagnoses. Overall, 277 (1.1%) patients had confirmed COVID-19 diagnosis, and 160 (0.7%) died, 9 with COVID-19.

Conclusion

Telemedicine resulted in low rates of ED visits and hospital admissions, suggesting positive impacts on healthcare utilization. Cardiovascular admissions were remarkably rare.

Introduction

On 26 February 2020, a new public-health concern emerged in Brazil, with the confirmation of the first coronavirus disease 2019 (COVID-19) case in the country. On 11 March 2020, the World Health Organization declared a pandemic, and by 20 August 2020, there were 3.47 million cases of COVID-19 in Brazil, with 111,443 deaths and 2.61 million recovered.¹,²

During the COVID-19 pandemic, adaptations of the basic structure for providing medical services have been proposed in response to the need to avoiding direct contact between providers and patients in non-emergent conditions.^3–5 In this setting, the use of telemedicine, boosted by other technology-based support systems, may drastically reduce the likelihood of transmission, as well as reducing staff contamination and thereby improving the availability of medical services. Moreover, telehealth services may also expand the capacity of multi-professional approaches, facilitating digital contact between professionals.³,⁴ A successful example of telemedicine utilisation to address the COVID-19 pandemic is the multimodal telemedicine network in Sichuan province, proven to be acceptable and effective in addition to generating improvements in overall health care.⁶

Telemedicine infrastructure differs widely between countries. Therefore, governmental and private incentives are needed to develop novel strategies aimed at further promoting safety of health professionals and serving the population at greatest risk. In 2016, the US Department of Health and Human Services estimated that >60% of US hospitals routinely used telehealth approaches.⁷ In response to the COVID-19 pandemic, federal US agencies promoted telemedicine through the relaxation of regulations and increased funding.⁷

UNIMED-BH is a non-profit private medical system (a work cooperative run by physicians) in the metropolitan area of Belo Horizonte, Southeast Brazil, currently with 5700 doctors and a network of 355 accredited services, covering approximately 1.28 million clients (21.5% of the population) through eight million yearly consultations, 24 million complementary exams and 150,000 hospitalisations. The development of the institutional plan for the COVID-19 pandemic started in December 2019, and a multimodal telemedicine support system was implemented in March 2020.⁸ We aimed to evaluate the impact of this large-scale COVID-19 telemedicine system on visits to emergency departments (ED) and all-cause and cardiovascular hospital admissions during the pandemic in Brazil.

Methods

Data sources

We conducted an observational three-arm study with prospective administrative data collection. Data analytical methods and study materials are available upon reasonable request from the corresponding author to other researchers for the purposes of reproducing the results or replicating the procedure.

The main pathway for the UNIMED-BH COVID-19 telemedicine system is an on-demand interactive app, for computers and mobile devices (www.unimedbh.com.br/coronavirus), launched on 18 March 2020 (Figure 1). From the opening screen, patients are directed to a list of questions about signs and symptoms of COVID-19 (Figure 2). Based on the number and/or severity of symptoms, moderate and severe cases are advised to schedule a teleconsultation through the app, with video interaction with an accredited physician trained in the management of COVID-19. During the appointment, the patient can be advised to stay at home and self-isolate with symptomatic treatment for flu syndrome, to schedule a face-to-face consultation, to present to the ED or to call an ambulance. Routine lab tests are recommended for moderate and severe cases that require visits to medical services. The coronavirus website⁸ additionally provides health information about COVID-19.

Figure 1.

Operational flow chart of the UNIMED-BH telemedicine system for coronavirus disease 2019 (COVID-19). ED: emergency department; EAS: emergency ambulance system (GMOV: Gerência de Atendimento móvel); VOIP: voice-over-IP. *Telemedicine: access through the online app, depending on the severity of flu-like symptoms checked. Patients can be advised to: (a) use conservative treatment, (b) self-isolate, (c) schedule a face-to-face appointment, (d) present to the ED or (e) call the EAS. Telemonitoring: regular calls for patients who reported respiratory symptoms, suspected or confirmed COVID-19 in any UNIMED-BH or accredited systems (including teleconsultation or ambulance systems); EAS: ambulance system, with Internet phone triage for ambulance activation and medical counselling in a triaging call centre, available for all clients.

Figure 2.

Screenshot of the UNIMED-BH mobile interactive coronavirus website, with questions about signs and symptoms of COVID-19. *Translations: nasal congestion, runny nose, shortness of breath, headache, sore throat, body aches, fever.

In a second pathway for monitoring disease progression, all patients with respiratory symptoms, suspected or confirmed COVID-19, contacting the UNIMED-BH services (including teleconsultation or ambulance systems) or any accredited services are included in an active telemonitoring system. Regular phone calls are provided in variable frequency, depending on disease severity and the profile of risk factors. Patients are also asked to self-monitor and report temperature and oxygen saturation whenever possible, but this information is not electronically fed into the app, and no specific devices are provided. In case of worsening symptoms, patients are advised to schedule face-to-face appointments through the app, to present to the ED or to access the ambulance triage. In the case of individuals with significant physical or neurological impairment, home visits by nurses and, if needed, physicians are provided.

In a third pathway, UNIMED-BH has an emergency ambulance system (EAS; Gerência de Atendimento Móvel (GMOV)), with a 24/7 triaging call centre for emergency calls and the management of the ambulance system (consisting of four advanced mobile intensive care units and 19 basic ambulances). Internet phone calls by clients are triaged by phone attendants and transferred to nurses or two physicians for counselling or ambulance activation when indicated. During the pandemic, EAS became part of the telemedicine system for patients with suspected COVID-19 for clarification, monitoring and urgent demands such as worsening symptoms and a need for medical care (Figure 1). In addition, telemedicine is being widely used for multiple sectors of UNIMED-BH, including EAS, for remote training on various aspects of COVID-19 management.

All data were exported from the UNIMED-BH central administrative database, hosted in the Management of Business Information Development (Gerência de Desenvolvimento de Informações para o Negócio (GDIN)), and entries were automatically de-identified by the system. All UNIMED-BH-owned and -accredited hospitals and clinics are connected to the administrative database, which is also linked to quality-of-care monitoring, audit and reimbursement. We collected demographic and clinical data from patients who accessed the teleconsultation and telemonitoring systems, and from those who accessed EAS with complaints classified by the physicians as suspected/confirmed COVID-19 or respiratory abnormalities. The study end points (clinical ED visits and hospital admissions) were prospectively recorded, and final diagnoses were assessed by the Diagnosis Related Groups (DRG) method.⁹ Codes associated with acute coronary syndromes (ACS), cardiovascular diseases (CVD), respiratory diseases and conditions requiring mechanical ventilation, according to the DRG diagnostic coding, were grouped separately (Supplemental Table S1). COVID-19 lab tests (real-time polymerase chain reaction and serology) and related deaths were also identified from the patient registries, as all institutions had a formal agreement to inform UNIMED-BH of new coronavirus diagnoses. As human subjects were not directly involved in the study, UNIMED-BH provided the Data Usage Agreement, and the research protocol was approved by the Institutional Review Board of Universidade Federal de Minas Gerais (no. 31095320.2.0000.5149).

Patient involvement

Patients and public were not involved in the design and conduct of this research.

Statistical analysis

Data were exported from the GDIN system as a Microsoft Excel file (Microsoft Corp., Redmond, WA) and imported to IBM SPSS Statistics v23.0 (IBM Corp., Armonk, NY) for statistical analysis. As an exploratory study, no pre-specified sample size calculation was performed, and we considered the total sample accessing the telemedicine system in its first 60 days. End points were collected for 14 days after the last inclusion (1 June 2020). Categorical variables, expressed as numbers and percentages, were compared between groups (telemonitoring, teleconsultation and EAS) using the chi-square test, whereas continuous data, expressed as the mean±standard deviation or median (interquartile range (IQR); 25–75%), were compared using one-way analysis of variance or the Kruskal–Wallis test, as appropriate. A two-tailed significance level of 0.05 was considered.

Results

In 60 days, the COVID-19 website was accessed 390,930 times (68.3% with smartphones, 31.0% with desktops and 0.9% with tablet computers), with a regular daily pattern (Figure 3), and the teleconsultation interactive menu was initiated 144,955 times. Of these, clinical questions were answered 81,711 (57.0%) times, and 52,155 (36.0%) processes were finalised and redirected to schedule teleconsultation. The age distribution of patients who reached this step is shown in Table 1. The mean access time was two minutes and 56 seconds.

Figure 3.

Number of daily unique accesses to the UNIMED-BH coronavirus website in the study period (March-May 2020).

Table 1.

Age distribution of patients who accessed the telemedicine system menu to schedule teleconsultations.

Age group (years)	n (%)
18–24	12,778 (24.5)
25–34	21,540 (41.3)
35–44	8449 (16.2)
45–54	4068 (7.8)
55–64	3077(5.9)
65+	2243 (4.3)
Total	52,155 (100)

Overall, 24,354 patients (1.9% of clients) actively participated in one of the telemedicine systems. Of these, 58.3% were women, and their median age was 36±18 years. The most frequently utilised service was telemonitoring (16,717; 69%), followed by teleconsultation (13,357; 55%) and EAS (687; 3%); 6,337 (26%) patients accessed or were included in more than one service. Detailed characteristics of the study groups are provided in Table 2.

Table 2.

Baseline characteristics and clinical outcomes of patients included in each of the three telemedicine support systems.

Variable	Total (N=24,354)	Telemonitoring (a) (N=16,717)	Teleconsultation (b) (N=13,357)	EAS (c) (N=687)	p-value
Age (years), M±SD	36±19	35±19	37±16	62±28	<0.001*,a
Sex (female), n (%)	14,210 (58.3)	9498 (56.8)	8236 (61.7)	411 (59.8)	<0.001*,b
Outcomes
ED visits, n (%)	4231 (17.4)	3296 (19.7)	2313 (17.3)	384 (55.9)	<0.001*,a
Number of ED visits/patient (M±SD)	1.3±0.7	1.3±0.7	1.3±0.7	1.4±1.2	0.033*,a
Hospital admissions, n (%)	1010 (4.1)	782 (4.7%)	318 (2.4)	388 (56.5)	<0.001*,c
Number of admissions/patient (M±SD)	1.1±0.5	1.1±0.4	1.1±0.5	1.2±0.7	0.002*,a
Hospital days, median (Q1–Q3)	6 (3–8)	6 (3–8)	4 (2–7)	8 (4–8)	0.04*,c
DRG diagnoses, n
Respiratory diseases	363	324	116	180	N/A
Requiring MV	44	35	3	29
CVD	44	29	8	19
ACS	5	4	0	1
Deaths, n (%)	160 (0.7)	120 (0.7)	9 (0.1)	90 (13.1)	<0.001*,c
COVID-19 lab diagnosis, n (%)	277 (1.1)	277 (1.7)	144 (1.1)	28 (4.1)	<0.001*,c

^ac>a=b.

^bc=a<b.

^cc>a>b.

*p<0.05.

ACS: acute coronary syndromes; CVD: cardiovascular diseases; DRG: Diagnosis Related Groups; ED: emergency department; EAS: emergency ambulance system; MV: mechanical ventilation; N/A: not applicable; Q1–Q3: quartiles 25–75%; SD: standard deviation.

A total of 21,371 teleconsultations were scheduled, and patients attended 17,334 (81%) of these (1.3 per patient). Among patients who attended at least one teleconsultation, their mean age was 37±16 years, and rates of ED visits and hospital admissions were, respectively, 17.3% (2313) and 2.3% (318). The median intervals between the teleconsultation and the first ED visit and hospital admission were, respectively, two (IQR 0–12) and seven (IQR 2–32) days, and 36.1% (6029) were included in telemonitoring for suspected respiratory symptoms. Nine (0.1%) patients died, one with confirmed COVID-19. For all patients included in telemonitoring for any period, their mean age was 35±19 years, 19.7% (3296) presented to the ED and 4.7% (782) were admitted to hospital. The mortality rate was 0.7% (120), eight with a COVID-19 diagnosis. Individuals utilising the EAS emergency calls were older (62±28 years), and the outcomes were considerably more frequent: 55.9% (384) presented to the ED, 56.5% (388) patients had at least one hospital admission and 90 (13.1%) died (five with confirmed COVID-19), of which 23 were home deaths, although none of these were with a COVID-19 diagnosis. Among those presenting to hospital or the ED, 435 were transported by ambulance. A total of 17.5% (120) of patients who accessed the EAS only received medical advice or counselling by phone. Table 2 provides detailed outcome data for each group.

In the total population, 4.1% (1010) had at least one hospital admission, and 107 had more than one admission, with a median of six (IQR 3–8) hospital days. Among these, 35.9% (363) had respiratory diseases according to the DRG coding (44 requiring mechanical ventilation), and 4.4% (44) had CVD (five ACS). Overall mortality in the study period was 0.7% (160 patients), with nine deaths associated with confirmed COVID-19. Other informed causes of death were respiratory diseases (13.8%), respiratory diseases requiring mechanical ventilation (11.3%), CVD (4.4%) and ACS (1.3%). Among patients who died, 52 (32.5%) had no recent hospital admissions and thus no DRG diagnosis. In total, 1.1% (277) of the patients had a confirmed COVID-19 lab diagnosis, and all of them were telemonitored for some period.

Discussion

Our data from one of the largest private health insurance companies in Brazil suggest that the implementation of a multimodal telemedicine system focused on the COVID-19 pandemic is feasible in the country. The adherence of clients was high in the short term, and considering the psychosocial stress resulting from the pandemic, rates of ED visits and hospital admissions were considerably low. Noticeably, the number of admissions associated with CVD and ACS was below expectations, although the state still experienced the ascending curve of COVID-19 incidence.

Although telemedicine was not broadly incorporated into medical regulations in Brazil before the COVID-19 pandemics, robust telehealth networks already existed in the country – noticeably in the public-health system held by universities – providing services such as tele-ECG, tele-retinography and tele-consulting between health providers.¹⁰ However, in response to the growing impact of the pandemic, the government issued law number 13,989 on 20 April 2020,¹¹ expanding the use of technology to other areas, such as teleconsultations and remote case monitoring. Anticipating the massive impact on private health systems, the planning of the UNIMED-BH COVID-19 telemedicine network started prior to this law, considering the complexity involved in the development and integration of multiple systems. Although this strategy has been widely recommended for expanding access to medical care and reducing patient and staff infection in the pandemic,^3–5 there is a paucity of data about practical results of telemedicine systems dedicated to COVID-19. For instance, in Western China, a multimodal network was successfully implemented for staff education, teleconsultations, phone and Internet clinical monitoring, adjustment of medical therapy and oxygen supplementation, as well as for remote acquisition of radiological imaging.⁶ However, the numerical impact on health-care utilisation was not reported in this study.

The UNIMED-BH strategy was favoured by factors related to the company’s pre-existent solutions for interaction with clients. First, a complex technology information system had already been developed for multiple purposes, including online scheduling of medical appointments, authorisation of procedures and patient/staff education. Furthermore, the majority of interactions with clients are already made through a patient-friendly mobile app, which was then adapted for the COVID-19 systems. Second, there was a pre-established programme of home care and monitoring for vulnerable patients, especially those with physical and/or neurological impairment.⁸ In this setting, technology-based solutions were well received by the covered population. Our choice to analyse the first eight weeks after the system’s deployment – as opposed to the peak of the pandemic – relied on the need to assess the population’s reaction to the early spread of COVID-19 in Brazil, which may have been influenced by fear of infection and extensive media coverage. As previously reported, in addition to the growing number of severe cases, a surge in cases reporting to the ED may be responsible for overcrowding and the collapse of the health system.¹² Conversely, some regions experienced a reduction in visits to the ED due to public concerns related to social-distancing policies.¹³ Despite the absence of comparable periods or similar baseline data, our findings suggest that multimodal telemedicine was associated with substantially low rates of ED and hospital admissions. This was remarkable for a population under telemonitoring – which partially overlaps with those utilising teleconsultations who were put under telemonitoring for some period. The less favourable outcomes for EAS, however, were possibly impacted by the high number of patients with chronic conditions and those under palliative care.

The decreased rates of admissions from our database, and especially the small number of patients with CVD diagnoses, must be interpreted cautiously in the light of trends observed in some countries. Several factors, including avoidance of medical care due to social distancing, psychological stress and concerns about becoming infected in the hospital, have led to a marked decrease in hospital admissions due to CVD.¹³,¹⁴ In Italy, a 58% increase in out-of-hospital cardiac arrests from 2019 to 2020¹⁵ paralleled a significant decrease in hospital admissions due to ACS.¹⁶ Similarly, an estimated 38–40% reduction in activations of cardiac catheterisation facilities due to ACS was observed early in the pandemics in the USA and Spain.^14–17 In Brazil, the civil registry database depicted a similar pattern, with increased home deaths and decreased deaths associated with acute cardiovascular events, contrasting with growing rates of unspecified cardiovascular causes (those without specific diagnosis).² In all these countries, the excess mortality in the period was remarkable,¹⁸ reinforcing the importance of telemedicine to provide contact with the health system during social isolation and also as a public awareness tool.³,⁴ Data suggest that this is being achieved in our population, given the high utilisation that followed the implementation of the multimodal strategy, but long-term assessment is warranted.

Low rates of COVID-19 lab diagnosis compared to other scenarios are also noticeable from our data. Currently, doubts remain about the ideal testing strategy, given the assumption that routine testing should only be implemented if results will lead to specific actions. This is even more debatable in low- and middle-income regions.¹⁹ In the Brazilian public-health system, the availability of tests was markedly low, especially at the beginning of the pandemic, and this shortage extended to private systems. Thus, the local policy was not to advise all symptomatic patients to undergo testing, which was restricted to those with moderate/severe symptoms requiring medical care, health-care personnel and individuals with high-risk or vulnerable home contacts.¹⁹ Self-isolation, however, was recommended for all suspected cases, and this was reinforced during the teleconsultation and telemonitoring interactions. This, in addition to the early phase of the pandemic, helps to explain our low number of confirmed cases and deaths associated with COVID-19. Nonetheless, this scenario seems to be changing as the pandemic evolves in the Belo Horizonte area.

Finally, the UNIMED strategy for responding to the COVID-10 pandemic combined interchangeable and collaborative approaches, with different systems supporting each other, in agreement with most position statements about telemedicine during the COVID-19 pandemic.^3–6,⁸ While teleconsultations connect providers with the covered population, minimising contamination and system overload, telemonitoring tracks the evolution of suspected and confirmed cases, reassuring isolation measures, and both are supported by a robust emergency system for orientation and ambulance support. The positive observations during the implementation phase may serve as examples for other health systems and even for public policies. Recently, the existing coronavirus app and the teleconsultation network were made available for the public-health system of the city of Belo Horizonte, an example of public–private collaboration to mitigate the effects of COVID-19 on the population.

Limitations

Our study has several limitations. First, it used a single-centre telemedicine system, which basically covers one Brazilian metropolitan area. Although the findings cannot be generalised, the size of the population (more than 1.28 million, 21.5% of the target population) and the vast area covered provide a representative insight into the broad effects of COVID-19 telemedicine support. Second, the analysis was performed in the first two months of the COVID-19 pandemic in a state spared in this phase due to social isolation policies. Thus, additional analyses at the peak of the COVID-19 outbreak, when health systems may be overloaded by a higher number of patients in need of medical care, are warranted in the near future. Third, the DRG was the only coding system applied for hospital admissions, and only the final main diagnoses were considered, limiting more detailed insights about underlying health conditions. Diagnoses for ED visits and causes of death for patients who died without admissions were also not available. In addition, COVID-19 is not specifically coded, and no inferences about COVID-19 exacerbating underlying diseases were possible. Nevertheless, DRG is widely validated for quality of care assessment and result-based reimbursement models, being the tool applied by the UNIMED system.⁹ Fourth, the medical recommendation and counselling for patients accessing the different systems was not recorded. Thus, it was not possible to infer if the individuals followed medical advice for staying at home or seeking emergency care. Lastly, it is not possible to compare our data to pre-pandemic periods, as this complex telemedicine system was developed for the COVID-19 pandemic. Despite these limitations, to the best of our knowledge, this is the largest report on a Latin American multimodal telemedicine system designed for the COVID-19 pandemic, and provides important data for delineation of technology-based support systems.

Conclusion

A multimodal telemedicine system resulted in low rates of ED visits and hospital admissions, suggesting a positive impact on large-scale health-care utilisation. The percentage of admissions due to CVD causes was low. As the results reflect the period before the peak of the pandemic in this region, further assessments are necessary to refine the ideal technology-based strategies to mitigate the impacts of COVID-19.

ORCID iDs

Bruno R Nascimento https://orcid.org/0000-0002-5586-774X

Luisa CC Brant https://orcid.org/0000-0002-7317-1367